What I’ve learned from leading data science teams

I have been leading teams for at least 12 years now, firstly teams of statisticians, then data scientists. I've been thinking about some of the key things I have learned over that time and I've outlined them below in no particular order. Do you relate to these, or are there other points I've missed?

Team wellbeing. To work productively, we need to be content (or at least, not unhappy!) in our jobs. I'm also a caring person so want those around me to be happy with what they're doing. To help this along, I ensure I discuss team member's interests and aims at work so that when we are looking at who should do what, people are broadly doing what they are interested in or something that they want to learn.

My wellbeing. Of course I care about my team so when issues at work or at home affect them, I have tried to be there to offer a shoulder to cry on or offer support. This has occasionally backfired on me in cases when people have relied on me too much to be their support, especially with an issue that I'm unable to do much about as a manager, and my wellbeing has been affected. I'm keen to not repeat such situations so I have adapted my approach to care, but also to enable them to find solutions to their problems.

Share rubbish work around. There are dull parts of every job so it is important to share these around, and be clear that they are being shared around so that nobody gets lumbered with them all. I once worked in a small team of three people where the major role was to type postcodes into Google and find out what was at that postcode. This is boring work and I remember one person not doing any of it because he knew it was too boring. Needless to say that this type of behaviour builds resentment in a team.

Share interesting work around. There are some pieces of research that just look cool before you start. Perhaps they involve an interesting new technique that people have been wanting to try out or the work is high profile. It's so important to share these around so that nobody feels they're getting a raw deal or that they're missing out on the fun stuff.

Share expertise. This is so important when managing data science teams where the techniques evolve so rapidly. I have also worked in teams where methods haven't changed in a long time but one (very competent) person leaves and others don't know how the method works or what it does.

As well as the obvious show and tells, encouraging team members to work on different things and learn from each other is valuable. No expertise should only be in one person.

Encourage customer centred thinking. This sounds really obvious but still sometimes requires explicit thinking about. For example, when I was leading a project a couple of team members suggested improvements to the accuracy of the model. But they hadn't actually checked with the customer whether they would prioritise this accuracy over other features such as the customer's deadlines.

Meet regularly about work and development with all team members individually, not just those who report directly to you. This is useful to understand what is really happening in the team and to understand team members' development needs. It's also helpful for understanding if there is unhappiness or tension in the team which may not always be reported by those working directly to you.

Build capability in those you work with. This is really important in data science and machine learning, where often customers don't understand what these can and can't do. For example one customer I worked with assumed these techniques could be used to interpret long contracts (sorry, not possible). I found it just as important to work with other data analysts or statisticians to build skills in data science, with some of my team working with and mentoring them.

I'm sure there are other things I have learned, but these are the key themes in my experience.